NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Ranji SR, Shetty K, Posley KA, et al. Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies (Vol. 6: Prevention of Healthcare–Associated Infections). Rockville (MD): Agency for Healthcare Research and Quality (US); 2007 Jan. (Technical Reviews, No. 9.6.)

Cover of Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies (Vol. 6: Prevention of Healthcare–Associated Infections)

Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies (Vol. 6: Prevention of Healthcare–Associated Infections).

Show details


Although many studies have been published documenting the effect of quality improvement initiatives on prevention of healthcare-associated infections, the published literature is of poor methodologic quality overall and does not consistently demonstrate the effectiveness of any specific strategy to either reduce infection rates or improve adherence to recommended preventive interventions. The available evidence does identify several promising strategies that merit more rigorous evaluation and may be appropriate for wider implementation. In the following sections, we will summarize our findings for prevention strategies for each target HAI, based on the few controlled studies and simple before-after studies with moderate to good methodologic quality.

Surgical Site Infection

Limited data (consisting of two interrupted time series67, 86 and one before-after study19) indicate that educational interventions coupled with audit and feedback may be effective at improving adherence to recommended strategies for SSI prevention, specifically appropriate antibiotic prophylaxis. Importantly, these strategies resulted in significant improvements in appropriate antibiotic timing, which has been documented to be deficient in U.S. hospitals. Clinician reminders may also improve perioperative antibiotic prophylaxis (one RCT, one CBA, one SBA), especially when incorporated into a computerized physician order entry system. No conclusion can be reached regarding the effectiveness of educational interventions alone on improving antibiotic prophylaxis practices. We also could not determine the effectiveness of QI strategies at promoting perioperative glucose control, perioperative normothermia, or decreasing operative site shaving, as very few studies reported data on these process measures.

We were also unable to determine any strategies effective at reducing SSI rates. Overall, SSI rates were statistically significantly reduced in five of 18 studies reporting this measure. In studies that did not have important methodologic flaws, surgical site infection rates were not consistently reduced, even when process measurements were improved. One study using an explicit “bundle” of interventions did improve process measures19 but not infection rates. Audit and feedback of SSI rates has been widely advocated, but the effect on surgical site infection rates is not clear. This strategy was evaluated in three multicenter studies (two ITS, one CBA66), with inconsistent results.

Central Line-Associated Bloodstream Infections

Two controlled studies,94, 99 one interrupted time series,96 and four simple before-after studies88, 89, 92, 93 of relatively good methodological quality used active educational interventions to significantly reduce the incidence of CLABSI. These interventions used demonstrations and self-study tutorials to improve adherence to preventive practices during catheter insertion. These educational interventions have been evaluated in teaching and non-teaching hospitals, and in U.S. and European institutions, increasing their generalizability. Two studies94, 96 used an explicit checklist to be filled out during line insertion, with nurses empowered to stop the procedure if all preventive interventions were not used, and documented marked reductions in CLABSI rates. This strategy may be worthy of wider implementation given its apparent success in a large population of ICUs in one study.96 We were unable to determine which QI strategies are effective at improving specific preventive interventions, as the studies documenting reduced CLABSI rates did not consistently report process measures.

Ventilator-Associated Pneumonia

Active educational interventions with use of a self-study module for ICU staff appear to be a promising strategy for reducing VAP rates, based on two SBA studies.107, 110 These studies used explicit clinical guidelines for preventing VAP, incorporating promotion of semirecumbent patient positioning and hand hygiene along with oral care, handling of ventilator condensate, and other interventions. No conclusion can be reached on the effectiveness of audit and feedback or other QI strategies on VAP rates. One SBA study effectively improved adherence to semirecumbent patient positioning using an educational- and reminder-based intervention.114

Catheter-Associated Urinary Tract Infection

Reminders to clinicians appear to be effective at reducing unnecessary catheter usage, primarily by reducing the duration of catheterization (two CBA studies123, 124 and two SBA studies119, 122). A key element of these studies was the use of an “automatic stop order” mandating discontinuation of the catheter after a specific time period (48 to 72 hours) unless the physician countermands the order. Three SBA studies using automatic stop orders were also associated with reduced CAUTI rates.119, 121, 122 Two of these studies121, 122 allowed nurses to remove catheters without a physician order if prespecified indications were met, an intervention worthy of future study. We could not determine the effect of other QI strategies on either infection rate or process measures. The safety of these interventions—i.e., the need for urethral re-catheterization—has not been adequately assessed; however, although re-catheterization is undoubtedly uncomfortable and inconvenient for patients, it is unlikely to lead to lasting harm. There is insufficient evidence supporting the utility of guidelines for catheter care.


The quality of included studies was poor. Across all four target HAIs, 52 of 64 included studies used a quasi-experimental, before-after design. Even within the limitations of this study design, most studies had poor internal validity, chiefly due to reporting data at only one time point before and after the intervention (33 of 52 studies). Whether the study reported infection rates or process measures (or both), it is very difficult to attribute an improvement in outcomes to the intervention on the basis of two data points. The baseline level of HAI rates were generally relatively high in most studies, and were above the pooled NNIS median for CLABSI, VAP, and UTI. While the NNIS data is not recommended for direct inter-hospital comparisons, the relatively high baseline rates raise the concern that the observed improvement (especially in before-after studies) could be due to regression to the mean, or even that institutions were motivated to study interventions to decrease HAI because of their unusually high infection rates. This may simply reflect the fact that many included studies were likely performed as documentation of ongoing quality improvement efforts, and not with a specific research agenda. Reporting of more time points of baseline data would provide assurance that an acute outbreak had not temporarily elevated baseline infection rates, and reporting of more data points after the intervention would provide greater assurance about the true intervention effect and its sustainability. Studies generally used NNIS definitions and used appropriate adjustment for device utilization, but given the problems with internal validity, the vast majority of studies cannot be considered generalizable.

We identified only one randomized controlled trial, eight controlled before-after trials, and three interrupted time series. These studies also exhibited problems with methodologic quality, especially among the CBA studies, in which the majority did not document the rationale for the selection of the control group. None of the controlled trials documented blinding of the outcomes assessors.

Reporting of infection rates without reporting accompanying process measures was common in our included studies. Of the 47 before-after studies reporting infection rates, only 24 reported data on the process measures targeted by the intervention. Accomplishing full adherence to recommended care can be difficult even in a clinical trial setting; a recent randomized trial of semirecumbent patient positioning found that the target elevation of the head of the bed (45 degrees) was not achieved in 85 percent of patients randomized to semirecumbent positioning,125 even with the presence of a dedicated research nurse assisting with intervention implementation. Given the inherent difficulties present in measuring HAI rates, and the lack of validated methods for inter-hospital comparisons of HAI rates, it is very important for process measures to be documented and reported. However, many of our included studies implemented an intervention with the intention of improving process measures, found a reduction in infection rates, and reported that the intervention must have been effective at improving the process, without actually documenting so. This is a particular concern in studies using passive interventions such as guideline dissemination and lectures. Past research has demonstrated that passive interventions are unlikely to achieve significant improvements in provider behavior,126 and thus it is unlikely that significant improvement in infection rates should occur as a result of such interventions if the appropriate process measures are not in fact improved.

Although several studies used an explicit clinical guideline for preventing HAIs (particularly the studies using educational interventions with self-study tutorials to target CLABSI and VAP), only two studies19, 113 directly assessed implementation of the “bundles” recommended by the IHI. The organization makes the point that setting the target of complete adherence to a bundle of processes “sets the bar high” and motivates overall system redesign rather than targeted single-process interventions,44 but there are no data to support this theoretically attractive claim. The IHI also recommends specific QI strategies for implementing the “bundles”, such as audit and feedback of infection rates and all-or-none measurements, and use of multidisciplinary rounds and setting daily patient goals for ICU patients.7 The very limited published data does not allow evaluation of the effectiveness of these strategies. The recommendations of the “100,000 Lives” campaign are being widely implemented in U.S. hospitals, providing an excellent opportunity for conducting higher-quality studies to determine effective implementation strategies.

In this review, the vast majority (30 of 39) of the studies reporting adherence to process measures reported a statistically significant improvement in adherence. This striking lack of reporting of negative results is highly likely to be a manifestation of publication bias. Although this trend was not as evident among studies reporting infection rates, 21 of 33 studies reporting CLABSI, VAP, or CAUTI rates found a statistically significant improvement. Thus, even the limited conclusions we are able to draw from the evidence may not be representative of overall experience with these strategies.

Other methodologic problems in the included studies are similar to those identified in previous volumes in this series. While most studies identified a baseline quality gap (generally an elevated HAI rate compared to benchmark standards), the majority did not specifically identify barriers to implementation of evidence-based practices or tailor their intervention to overcome barriers. Most studies provided few details on the intervention, particularly with regards to the intensity of the intervention and its reach (the extent to which those targeted by the intervention actually received it). Most studies did not state if other QI interventions were underway simultaneously. The median length of follow up was approximately one year across all studies, which is likely too short to confirm a sustained improvement in either infection rates or process measures.

Because of the limited number of controlled trials, we were unable to perform any quantitative analyses such as median effects analysis. We are thus unable to obtain any estimate of the magnitude of the effect that hospitals implementing these strategies may hope to achieve. Very few studies reported on any potential adverse effects of the intervention, and no high-quality studies assessed the cost-benefit of the intervention.


Due to the extensive limitations in the primary data outlined above, we are not able to make any firm recommendations for organizations seeking to implement quality improvement interventions to reduce healthcare-associated infections. Quality improvement efforts in infection control are active, and thus we make the following recommendations for further study in this area.


Preliminary data indicates that several strategies are worthy of future study.

Based on limited evidence, the following quality improvement strategies to reduce healthcare-associated infections could be considered for wider implementation, if higher-quality studies confirm their effectiveness:

  • Printed or computer-based reminders with use of automatic stop orders to reduce unnecessary urethral catheterization. This is the only strategy we identified that is supported by more than one controlled study.
  • Printed or computer-based reminders for improving adherence to recommendations for timing and duration of surgical antibiotic prophylaxis;
  • Staff education using interactive tutorials (including video and web-based) and checklists, to improve adherence to insertion practices for placement of central venous catheters;
  • Staff education, including use of interactive tutorials, to improve adherence to preventive interventions to prevent ventilator-associated pneumonia.

Higher quality studies of QI strategies to implement preventive interventions are urgently needed.

Given the prevalence of HAIs and the associated morbidity and mortality, there is a great need for information on how to improve adherence to preventive interventions. We recognize that conducting controlled trials may not be practical for many investigators for cost and feasibility reasons; as well, it may be unethical to randomize subjects to receive or not receive preventive interventions. If performing a controlled trial is impractical, investigators should perform interrupted time series analyses to demonstrate that a QI intervention is truly effective. At least three time points of data should be reported before and after a clearly defined intervention time period, and a formal ITS statistical analysis should be conducted. Studies should ideally report the effect of the intervention on both process measures (adherence to recommended preventive interventions) and infection rates. Studies should also document if other QI efforts were underway simultaneously, and use standardized definitions for measuring and reporting HAIs. Formal evaluation of the cost-benefit of the intervention should be conducted, and adverse events should be documented.

We acknowledge that even conducting higher-quality nonrandomized trials may pose logistical challenges. Monitoring infection rates over a sustained period of time requires adequate infection control resources, which may not currently be in place at many U.S. hospitals.127, 128 In an era of public reporting of infection rates, hospitals may feel pressured to respond quickly to increases in measured infection rates, even if these rates may not be directly tied to poor adherence to preventive practices. Measuring adherence to preventive interventions is resource intensive,129 and requires additional trained personnel. If public reporting of infection rates becomes more widespread, as seems likely, investment in infection control resources will be necessary in order to adequately monitor infection rates and process measures and continue study of implementation of effective preventive measures.

We also believe that there are potential cross-cutting opportunities, in which interventions proven to work for one target should be considered for others. For example, empowering nurses to remove urinary catheters when patients met prespecified criteria, when coupled with earlier studies that demonstrated the value of respiratory-therapist-driven ventilator weaning algorithms,130 may point the way to more interventions managed by nonphysician providers, working under carefully crafted protocols. Another example is empowering nurses to stop central venous catheter insertion if a checklist of preventive interventions is not followed; empowering nonphysician providers in this fashion could be applied to preventive interventions for many HAIs.

We emphasize that the level of evidence supporting these strategies is below that supporting the recommendations made in prior volumes of this series.47, 48 Thus, we are unable to recommend strongly that these QI strategies be more widely implemented, as—in addition to the poor quality of supporting evidence—the potential adverse consequences and cost-benefit of these strategies has not been assessed. Wider implementation of these measures (most of which do not appear to be complex or costly) should be feasible, but should only be performed if a formal plan for evaluating their effectiveness is in place.

The lack of strong evidence supporting use of specific QI strategies should not be taken to mean that ongoing QI efforts in HAI prevention have been uniformly unsuccessful, or that current strategies should not be continued. Population-level data from Europe shows that HAI incidence can be reduced through use of infection control surveillance,131 and some institutions have documented sustained reductions in specific HAI.105, 106 The mechanisms behind these successes and the best means of translating them into other settings remain to be determined. High-quality evidence exists to support many preventive interventions that are very effective at reducing HAI incidence. Given the huge toll in human lives, antibiotic use (leading to more resistant infections) and costs caused by hospital-acquired infections, efforts to better understand how to implement these interventions should be prioritized.

PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (785K)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...